Nowadays, the amount of data that can be processed daily has reached quite big dimensions with the developments in the big data. The fact that a large part of this data consists of text data has made the studies in the field of text processing very important and popular. However, when the studies in this area are examined, it has been observed that while various studies are carried out for many World languages, especially English, there are no desired number of studies conducted in Turkish. Therefore, a large corpus of Turkish texts was created using the Beautiful Soup library, one of the python environment libraries. In this study, CBOW and Skip-Gram algorithms from Word2Vec model algorithms and Glove model were used where each word is represented with a vector in the vector space. In this study, a model which consists of Word2Vec method and Turkish words and tries to detect the semantic relations between these words has been developed and the performance and training times have been compared with other models. In addition, another contribution of this study to improve the performance of the model to create a list of stop words for Turkish. With this model, it is aimed to achieve a higher classification performance especially for Turkish text classification problems. After analyzing the model formed within the scope of this study, it was detected that it showed a very successful performance when close words were examined. The dataset and word vectors will be shared with the public to provide contributions to Turkish studies.
Primary Language | Turkish |
---|---|
Journal Section | Articles |
Authors | |
Publication Date | June 30, 2019 |
Published in Issue | Year 2019 Volume: 34 Issue: 2 |